Subseasonal Predictions of Tropical Cyclone Occurrence and ACE in the S2S Dataset

Abstract

Probabilistic tropical cyclone (TC) occurrence, at lead times of week 1–4, in the Subseasonal to Seasonal (S2S) dataset are examined here. Forecasts are defined over 15° in latitude × 20° in longitude regions, and the prediction skill is measured using the Brier skill score with reference to climatological reference forecasts. Two types of reference forecasts are used: a seasonally constant one and a seasonally varying one, with the latter used for forecasts of anomalies from the seasonal climatology. Models from the European Centre for Medium-Range Weather Forecasts (ECMWF), Australian Bureau of Meteorology, and Météo-France/Centre National de Recherche Météorologiques have skill in predicting TC occurrence four weeks in advance. In contrast, only the ECMWF model is skillful in predicting the anomaly of TC occurrence beyond one week. Errors in genesis prediction largely limit models’ skill in predicting TC occurrence. Three calibration techniques, removing the mean genesis and occurrence forecast biases, and a linear regression method, are explored here. The linear regression method performs the best and guarantees a higher skill score when applied to the in-sample dataset. However, when applied to the out-of-sample data, especially in areas where the TC sample size is small, it may reduce the models’ prediction skill. Generally speaking, the S2S models are more skillful in predicting TC occurrence during favorable Madden–Julian oscillation phases. Last, we also report accumulated cyclone energy predictions skill using the ranked probability skill score.

A Hybrid Dynamical‐Statistical Model for Advancing Subseasonal Tropical Cyclone Prediction Over the Western North Pacific

“…The recurving trajectories of TCs approaching Japan and the Korean Peninsula are also predicted at a long lead (Figures 4c and 4d). The recent study of Lee et al (2020) found that some S2S models are skillful in predicting regional TCF (in every 15° × 20° grid cell) at a long lead time out to 4 weeks based on the assessment of Brier skill score. Because the definition of track density based on the method of track‐pattern‐based model (Chu et al, 2010) is different from the probability of TCF of Lee et al (2020), it is not yet clear whether our finer resolution (5° × 5°) prediction of TC density outperforms the dynamical prediction of the S2S models.…”

“…Subseasonal prediction of TC activity has been challenging, while the state-of-the-art dynamical prediction systems show skills in the probabilistic predictions of TC occurrence anomalies with positive Brier skill scores at forecast leads up to 2-4 weeks over some basins (Gregory et al, 2020;Lee et al, 2018Lee et al, , 2020. Compared to the probabilistic prediction, the deterministic cyclogenesis (for both the location and time of TC genesis) prediction at the subseasonal timescale is much more difficult (Jiang et al, 2018).…”

“…provide information on the uncertainty of the forecasts (Gregory et al, 2020;Lee et al, 2018Lee et al, , 2020. We are working with operational centers in TC-prone areas of China to advance this work.…”

“…In this respect, the S2S project provides a vital data set for further establishing a real‐time TC prediction system based on our hybrid model prediction approach. Moreover, the ensemble members of the S2S prediction system can be efficiently used for probabilistic TC predictions, which provide information on the uncertainty of the forecasts (Gregory et al, 2020; Lee et al, 2018, 2020). We are working with operational centers in TC‐prone areas of China to advance this work.…”

“…Using the subseasonal‐to‐seasonal (S2S) prediction data set (Vitart et al, 2017), Lee et al (2018) documented that only a few of the six S2S models showed skills for TC genesis prediction with the 2‐week (8–14 days) forecasts. Gregory et al (2020) and Lee et al (2020) extended the work of Lee et al (2018) to assess prediction skills of probabilistic TC occurrence at a regional scale (15° latitude × 20° longitude) over different basins. They concluded that the ECMWF model and the prediction systems from Australia and France show good skills in predicting the probability of TC occurrence out to 3–4 weeks in advance.…”